隐式等权重粒子滤波在高维准地转模式中的特性研究  

Implicit equal-weights particle filter in a high-dimensional quasi-geostrophic model

作　　者：陈妍 朱孟斌 张卫民[1] 曹小群[1] CHEN Yan;ZHU Mengbin;ZHANG Weimin;CAO Xiaoqun(College of Meteorology and Oceanology, National University of Defense Technology, Changsha, 410073, China;College of Computer, National University of Defense Technology, Changsha, 410073, China)

机构地区：[1]国防科技大学气象海洋学院,长沙410073 [2]国防科技大学计算机学院,长沙410073

出　　处：《气象学报》2018年第2期255-265,共11页Acta Meteorologica Sinica

基　　金：国家自然科学基金项目(41375113;41475094和41675097)

摘　　要：粒子滤波自从被引入资料同化领域以来,对于高维系统存在的粒子衰退问题一直困扰着资料同化领域的研究。隐式等权重粒子滤波(Implicit Equal-Weights Particle Filter,IEWPF)通过在高维的状态空间维数的前提下,隐式从每个粒子都具有特殊协方差的提议密度中进行采样,构建等权重的粒子集合,从而解决高维系统的粒子衰退问题。通过在高维准地转模式中应用IEWPF方法,验证了IEWPF的系统一致性和资料同化效果。通过对水平动能谱的检验,验证了IEWPF可以保持系统的原始平衡特性。通过IEWPF与等权重粒子滤波(Equivalent Weights Particle Filter,EWPF)的对比试验发现,两者的资料同化分析场非常接近,但在运行效率上,IEWPF远优于EWPF。同时,IEWPF也为解决一系列的资料同化问题,比如参数估计,提供了新的解决途径。Implicit equal-weights particle filter（IEWPF）has beaten the ＂curse of dimensionality＂ in high-dimensional non-linear geophysical models,which has plagued the particle filters community for decades.The IEWPF draws samples from the proposal density with different covariances for each particle and construct equal-weights for all the particles.The IEWPF is implemented in a high-dimensional two-layer quasi-geostrophic model to check its features and properties.The experiment results show that the new particle filter method has a very good consistency and yields excellent data assimilation results.The horizontal kinetic energy spectrum test indicates that the new method can keep the original balance of the model variables.The comparison between IEWPF and equivalent weights particle filter（EWPF）reveals that IEWPF has a better computation cost than EWPF,while both of them are very good when compared with the analyses.Therefore,IEWPF opens a door for solving the advanced data assimilation problems such as parametrization,etc.

关 键 词：隐式等权重粒子滤波 维数灾难 提议密度 高维准地转模式 

分 类 号：P456.7[天文地球—大气科学及气象学]